summaryrefslogtreecommitdiff
path: root/tests/intel/xe_dma_buf_sync.c
blob: f0d3f0d549379eecce9e756a45fd39b5ad90ce22 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
// SPDX-License-Identifier: MIT
/*
 * Copyright © 2021 Intel Corporation
 */

/**
 * TEST: Check dmabuf functionality
 * Category: Core
 * Mega feature: General Core features
 * Sub-category: dmabuf
 * Test category: functionality test
 */

#include "igt.h"
#include "lib/igt_syncobj.h"
#include "lib/intel_reg.h"
#include "xe_drm.h"

#include "xe/xe_ioctl.h"
#include "xe/xe_query.h"
#include "xe/xe_spin.h"
#include <string.h>
#include <linux/dma-buf.h>
#include <poll.h>

#define MAX_N_BO	16
#define N_FD		2

#define READ_SYNC	(0x1 << 0)

struct igt_dma_buf_sync_file {
	__u32 flags;
	__s32 fd;
};

#define IGT_DMA_BUF_IOCTL_EXPORT_SYNC_FILE \
	_IOWR(DMA_BUF_BASE, 2, struct igt_dma_buf_sync_file)

static int dmabuf_export_sync_file(int dmabuf, uint32_t flags)
{
	struct igt_dma_buf_sync_file arg;

	arg.flags = flags;
	arg.fd = -1;
	do_ioctl(dmabuf, IGT_DMA_BUF_IOCTL_EXPORT_SYNC_FILE, &arg);

	return arg.fd;
}

static bool dmabuf_busy(int dmabuf, uint32_t flags)
{
	struct pollfd pfd = { .fd = dmabuf };

	/* If DMA_BUF_SYNC_WRITE is set, we don't want to set POLLIN or
	 * else poll() may return a non-zero value if there are only read
	 * fences because POLLIN is ready even if POLLOUT isn't.
	 */
	if (flags & DMA_BUF_SYNC_WRITE)
		pfd.events |= POLLOUT;
	else if (flags & DMA_BUF_SYNC_READ)
		pfd.events |= POLLIN;

	return poll(&pfd, 1, 0) == 0;
}

static bool sync_file_busy(int sync_file)
{
	struct pollfd pfd = { .fd = sync_file, .events = POLLIN };
	return poll(&pfd, 1, 0) == 0;
}

/**
 * SUBTEST: export-dma-buf-once
 * Description: Test exporting a sync file from a dma-buf
 * Functionality: export
 *
 * SUBTEST: export-dma-buf-once-read-sync
 * Description: Test export prime BO as sync file and verify business
 * Functionality: export
 *
 * SUBTEST: export-dma-buf-many
 * Description: Test exporting many sync files from a dma-buf
 * Functionality: export
 *
 * SUBTEST: export-dma-buf-many-read-sync
 * Description: Test export many prime BO as sync file and verify business
 * Functionality: export
 */

static void
test_export_dma_buf(struct drm_xe_engine_class_instance *hwe0,
		    struct drm_xe_engine_class_instance *hwe1,
		    int n_bo, int flags)
{
	uint64_t addr = 0x1a0000, base_addr = 0x1a0000;
	int fd[N_FD];
	uint32_t bo[MAX_N_BO];
	int dma_buf_fd[MAX_N_BO];
	uint32_t import_bo[MAX_N_BO];
	uint32_t vm[N_FD];
	uint32_t exec_queue[N_FD];
	size_t bo_size;
	struct {
		struct xe_spin spin;
		uint32_t batch[16];
		uint64_t pad;
		uint32_t data;
	} *data [MAX_N_BO];
	int i;

	igt_assert(n_bo <= MAX_N_BO);

	for (i = 0; i < N_FD; ++i) {
		fd[i] = drm_open_driver(DRIVER_XE);
		vm[i] = xe_vm_create(fd[i], 0, 0);
		exec_queue[i] = xe_exec_queue_create(fd[i], vm[i], !i ? hwe0 : hwe1, 0);
	}

	bo_size = sizeof(*data[0]) * N_FD;
	bo_size = xe_bb_size(fd[0], bo_size);
	for (i = 0; i < n_bo; ++i) {
		bo[i] = xe_bo_create(fd[0], 0, bo_size,
				     vram_if_possible(fd[0], hwe0->gt_id),
				     DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
		dma_buf_fd[i] = prime_handle_to_fd(fd[0], bo[i]);
		import_bo[i] = prime_fd_to_handle(fd[1], dma_buf_fd[i]);

		if (i & 1)
			data[i] = xe_bo_map(fd[1], import_bo[i], bo_size);
		else
			data[i] = xe_bo_map(fd[0], bo[i], bo_size);
		memset(data[i], 0, bo_size);

		xe_vm_bind_sync(fd[0], vm[0], bo[i], 0, addr, bo_size);
		xe_vm_bind_sync(fd[1], vm[1], import_bo[i], 0, addr, bo_size);
		addr += bo_size;
	}
	addr = base_addr;

	for (i = 0; i < n_bo; ++i) {
		uint64_t batch_offset = (char *)&data[i]->batch -
			(char *)data[i];
		uint64_t batch_addr = addr + batch_offset;
		uint64_t sdi_offset = (char *)&data[i]->data - (char *)data[i];
		uint64_t sdi_addr = addr + sdi_offset;
		uint64_t spin_offset = (char *)&data[i]->spin - (char *)data[i];
		struct drm_xe_sync sync[2] = {
			{ .type = DRM_XE_SYNC_TYPE_SYNCOBJ, },
			{ .type = DRM_XE_SYNC_TYPE_SYNCOBJ, .flags = DRM_XE_SYNC_FLAG_SIGNAL, },
		};
		struct drm_xe_exec exec = {
			.num_batch_buffer = 1,
			.syncs = to_user_pointer(sync),
		};
		struct xe_spin_opts spin_opts = { .addr = addr + spin_offset, .preempt = true };
		uint32_t syncobj;
		int b = 0;
		int sync_fd;

		/* Write spinner on FD[0] */
		xe_spin_init(&data[i]->spin, &spin_opts);
		exec.exec_queue_id = exec_queue[0];
		exec.address = spin_opts.addr;
		xe_exec(fd[0], &exec);

		/* Export prime BO as sync file and veify business */
		if (flags & READ_SYNC)
			sync_fd = dmabuf_export_sync_file(dma_buf_fd[i],
							  DMA_BUF_SYNC_READ);
		else
			sync_fd = dmabuf_export_sync_file(dma_buf_fd[i],
							  DMA_BUF_SYNC_WRITE);
		xe_spin_wait_started(&data[i]->spin);
		igt_assert(sync_file_busy(sync_fd));
		igt_assert(dmabuf_busy(dma_buf_fd[i], DMA_BUF_SYNC_READ));

		/* Convert sync file to syncobj */
		syncobj = syncobj_create(fd[1], 0);
		syncobj_import_sync_file(fd[1], syncobj, sync_fd);

		/* Do an exec with syncobj as in fence on FD[1] */
		data[i]->batch[b++] = MI_STORE_DWORD_IMM_GEN4;
		data[i]->batch[b++] = sdi_addr;
		data[i]->batch[b++] = sdi_addr >> 32;
		data[i]->batch[b++] = 0xc0ffee;
		data[i]->batch[b++] = MI_BATCH_BUFFER_END;
		igt_assert(b <= ARRAY_SIZE(data[i]->batch));
		sync[0].handle = syncobj;
		sync[1].handle = syncobj_create(fd[1], 0);
		exec.exec_queue_id = exec_queue[1];
		exec.address = batch_addr;
		exec.num_syncs = 2;
		xe_exec(fd[1], &exec);

		/* Verify exec blocked on spinner / prime BO */
		usleep(5000);
		igt_assert(!syncobj_wait(fd[1], &sync[1].handle, 1, 1, 0,
					 NULL));
		igt_assert_eq(data[i]->data, 0x0);

		/* End spinner and verify exec complete */
		xe_spin_end(&data[i]->spin);
		igt_assert(syncobj_wait(fd[1], &sync[1].handle, 1, INT64_MAX,
					0, NULL));
		igt_assert_eq(data[i]->data, 0xc0ffee);

		/* Clean up */
		syncobj_destroy(fd[1], sync[0].handle);
		syncobj_destroy(fd[1], sync[1].handle);
		close(sync_fd);
		addr += bo_size;
	}

	for (i = 0; i < n_bo; ++i) {
		munmap(data[i], bo_size);
		gem_close(fd[0], bo[i]);
		close(dma_buf_fd[i]);
	}

	for (i = 0; i < N_FD; ++i)
		drm_close_driver(fd[i]);
}

igt_main
{
	struct drm_xe_engine_class_instance *hwe, *hwe0 = NULL, *hwe1;
	int fd;

	igt_fixture {
		fd = drm_open_driver(DRIVER_XE);

		xe_for_each_engine(fd, hwe)
			if (hwe0 == NULL) {
				hwe0 = hwe;
			} else {
				hwe1 = hwe;
				break;
			}
	}

	igt_subtest("export-dma-buf-once")
		test_export_dma_buf(hwe0, hwe1, 1, 0);

	igt_subtest("export-dma-buf-many")
		test_export_dma_buf(hwe0, hwe1, 16, 0);

	igt_subtest("export-dma-buf-once-read-sync")
		test_export_dma_buf(hwe0, hwe1, 1, READ_SYNC);

	igt_subtest("export-dma-buf-many-read-sync")
		test_export_dma_buf(hwe0, hwe1, 16, READ_SYNC);

	igt_fixture
		drm_close_driver(fd);
}